Domains
Forms
Mechanism
Breaks down how the skill works and what it produces, so you can quickly judge whether it fits your scenario and value.
When raw data needs conversion into training format for SFT or preference training, after finetuning-method-selection routed here. Typical context: preparing JSONL, applying chat templates, configuring packing, generating synthetic data, or writing a dataset card.
A formatted, packed, validated JSONL dataset plus a complete dataset card — the Phase 2 artifact /finetune checks before launching training.
Use Cases
A team collected 5,000 human demo conversations and wants to SFT a support model. Using this skill, they pick the ChatML messages format, apply the model's chat template before packing, mask loss to assistant turns, pack sequences to save compute, and write a dataset card with provenance and dedup method. The validated JSONL and card pass the `/finetune` gate, avoiding silent template mismatches and padding waste.
A researcher generates 8,000 synthetic math prompts via Magpie but needs DPO pairs. The skill guides keeping ≥25% real replay rows, building preference pairs with `preference-optimization`, applying templates, and inspecting packed sequences. The dataset card documents synthetic/real ratio and packing config. This prevents model collapse and ensures the Phase 2 artifact is training-ready.
Skill Relationships
Dependency relationships read as "the upper tier points to the lower tier." The current Skill sits in the middle tier — above are Skills that depend on it, below are Skills it depends on.
Tier 1 · These Skills Use Me
Tier 2 · Current Skill
Tier 3 · I Use These Skills
Browse this skill's relationships within its skillset. Click a node to switch the side panel; use the search box to jump to any skill.
Skill File
